The Agent Skills Directory

[SAFE]: The skill provides legitimate guidance on hardening E2E tests based on established principles (independence, mocking, and flaky prevention). It includes warnings about environment variable leakage and emphasizes proper architectural boundaries.
[NO_CODE]: The skill is composed of instructional Markdown files and configuration JSON. It does not ship with executable scripts or binaries that could perform hidden actions.
[COMMAND_EXECUTION]: The skill recommends using standard developer tools such as grep and npx playwright test. These commands are used strictly for identifying anti-patterns in code and verifying test stability via burn-in procedures.
[PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection because it instructs the agent to read and process user-controlled test files and source code. This is classified as a low-risk surface inherent to code-editing skills.
Ingestion points: User-provided Playwright test specifications and application source code files (e.g., path/to/tests).
Boundary markers: Not explicitly defined in the instructions; the agent treats file content as source code to be refactored.
Capability inventory: The agent can execute shell commands (e.g., grep, npx playwright test), read files, and write modifications back to the filesystem.
Sanitization: The skill lacks automated sanitization for ingested content, but it explicitly instructs the agent to look for and prevent the exposure of sensitive internal environment variables.

e2e-test-hardener